Edinburgh Research Explorer · 2019-06-21 · May/June 2019, 6(3) e0106-19.2019 eNeuro.org. NMDA...

Edinburgh Research Explorer

Slow NMDA-Mediated Excitation Accelerates Offset-ResponseLatencies Generated via a Post-Inhibitory Rebound Mechanism

Citation for published version:Rajaram, E, Kaltenbach, C, Fischl, MJ, Mrowka, L, Alexandrova, O, Grothe, B, Hennig, MH & Kopp-Scheinpflug, C 2019, 'Slow NMDA-Mediated Excitation Accelerates Offset-Response Latencies Generatedvia a Post-Inhibitory Rebound Mechanism', eNeuro, vol. 6, no. 3. https://doi.org/10.1523/ENEURO.0106-19.2019

Digital Object Identifier (DOI):10.1523/ENEURO.0106-19.2019

Link:Link to publication record in Edinburgh Research Explorer

Document Version:Publisher's PDF, also known as Version of record

Published In:eNeuro

Publisher Rights Statement:Copyright © 2019 Rajaram et al.This is an open-access article distributed under the terms of the CreativeCommons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction inany medium provided that the original work is properly attributed.

General rightsCopyright for the publications made accessible via the Edinburgh Research Explorer is retained by the author(s)and / or other copyright owners and it is a condition of accessing these publications that users recognise andabide by the legal requirements associated with these rights.

Take down policyThe University of Edinburgh has made every reasonable effort to ensure that Edinburgh Research Explorercontent complies with UK legislation. If you believe that the public display of this file breaches copyright pleasecontact [email protected] providing details, and we will remove access to the work immediately andinvestigate your claim.

Download date: 01. Jun. 2020

https://www.research.ed.ac.uk/portal/en/persons/matthias-hennig(60ea63ba-9e12-4fc5-81c3-16ff526284f3).html

https://www.research.ed.ac.uk/portal/en/publications/slow-nmdamediated-excitation-accelerates-offsetresponse-latencies-generated-via-a-postinhibitory-rebound-mechanism(7e0c5226-d4ca-45b9-8e5c-34e189cc32df).html


https://doi.org/10.1523/ENEURO.0106-19.2019




Sensory and Motor Systems

Slow NMDA-Mediated Excitation AcceleratesOffset-Response Latencies Generated via a Post-Inhibitory Rebound MechanismEzhilarasan Rajaram,1,2 Carina Kaltenbach,3 Matthew J. Fischl,1,� Leander Mrowka,1 Olga Alexandrova,1

Benedikt Grothe,1 Matthias H. Hennig,3 and Conny Kopp-Scheinpflug1


1Division of Neurobiology, Department Biology II, Ludwig-Maximilians-University Munich, Planegg-Martinsried 82152,Germany, 2Graduate School of Systemic Neurosciences, Ludwig-Maximilians-University Munich, Planegg-Martinsried82152, Germany, and 3Institute for Adaptive and Neural Computation, School of Informatics, University of Edinburgh,Edinburgh EH8 9AB, United Kingdom

AbstractIn neural circuits, action potentials (spikes) are conventionally caused by excitatory inputs whereas inhibitory inputsreduce or modulate neuronal excitability. We previously showed that neurons in the superior paraolivary nucleus (SPN)require solely synaptic inhibition to generate their hallmark offset response, a burst of spikes at the end of a soundstimulus, via a post-inhibitory rebound mechanism. In addition SPN neurons receive excitatory inputs, but theirfunctional significance is not yet known. Here we used mice of both sexes to demonstrate that in SPN neurons, theclassical roles for excitation and inhibition are switched, with inhibitory inputs driving spike firing and excitatory inputsmodulating this response. Hodgkin–Huxley modeling suggests that a slow, NMDA receptor (NMDAR)-mediatedexcitation would accelerate the offset response. We find corroborating evidence from in vitro and in vivo recordingsthat lack of excitation prolonged offset-response latencies and rendered them more variable to changing soundintensity levels. Our results reveal an unsuspected function for slow excitation in improving the timing of post-inhibitoryrebound firing even when the firing itself does not depend on excitation. This shows the auditory system employshighly specialized mechanisms to encode timing-sensitive features of sound offsets which are crucial for sound-duration encoding and have profound biological importance for encoding the temporal structure of speech.

Key words: auditory development; duration encoding; gap-detection; level-tolerance; sound-offset encoding;superior paraolivary nucleus

IntroductionThe relative strength and temporal interaction of excit-

atory and inhibitory synaptic inputs determine neuronal

temporal firing patterns in many parts of the brain includ-ing the auditory pathway (Wehr and Zador, 2003; Sunet al., 2010; Denève and Machens, 2016). Acoustic pat-

Received March 20, 2019; accepted May 2, 2019; First published May 31,2019.The authors declare no competing financial interests.

Author contributions: E.R., C.K., M.J.F., L.M., O.A., and C.K.-S. performedresearch; E.R., C.K., M.J.F., O.A., and C.K.-S. analyzed data; E.R., B.G.,M.H.H., and C.K.-S. wrote the paper; M.H.H. and C.K.-S. designed research.

Significance Statement

Temporal features like sound-onset and sound-offset responses are crucial for acoustic pattern recognitionand vocal communication. In contrast to onset responses, offset responses can be generated solely frominhibition via post-inhibitory rebounds. Here, we demonstrate that excitatory inputs are neverthelesspresent at offset-encoding neurons of the auditory brainstem where they serve to shorten the responselatency of the offset response and stabilize the offset latency against changes in stimulus level.

New Research

May/June 2019, 6(3) e0106-19.2019 1–20


tern recognition and vocal communication require preciseprocessing of the temporal structure of sounds, such asthe distinct detection of onsets and offsets, which areencoded by two dissociable channels in the auditorypathway (Anderson and Linden, 2016; Gómez-Álvarezet al., 2018; Kopp-Scheinpflug et al., 2018).

In previous studies of the central auditory system,sound-offset responses were often ignored (Kopp-Scheinpflug et al., 2018). While sound offsets are typicallyless abrupt (Cavaco and Lewicki, 2007) and more easilyobscured by reverberation than onsets, and perceptuallyless prominent than onsets (Phillips et al., 2002; Deneuxet al., 2016; Sohoglu and Chait, 2016), this view is chang-ing with the discovery of neurons dedicated to detectingoffsets. Recent studies suggest that neural representa-tions of sound transients, such as offsets, are importantfor speech perception (Eggermont, 2015), are implicatedin auditory dysfunction in brain disorders (Felix et al.,2018) and appear to be involved in short-term memoryformation during auditory task performance (Elguedaet al., 2019). A previous study showed that neurons re-sponding to sounds in a sustained fashion do not encodethe end of a stimulus as reliably as those which respondspecifically to the offset (Kopp-Scheinpflug et al., 2011).Detection of the end of a sound stimulus using the lastsound-evoked spike in a neuron with a sustained re-sponse is highly variable, mainly due to cellular and syn-aptic properties changing during stimulus time. Atstimulus onset, a neuron is in resting conditions andtypically highly excitable, which allows fast and precisespiking. After a period of tonic activity, synaptic depres-sion and activation of voltage-gated conductances re-duce excitability, leading to less precise responses (cf.Kopp-Scheinpflug et al., 2011, their Fig. 7). Thus, a carefulexamination of auditory offset responses will aid in under-standing their underlying coding mechanisms, both ofwhich are prerequisites to study their behavioral rele-vance.

Sound-evoked offset responses have been reported inall processing stages of the ascending auditory pathwayfrom brainstem to cortex (Kopp-Scheinpflug et al., 2018).Offset responses in higher areas such as the auditorycortex seem to be largely inherited from subcortical struc-tures via nonoverlapping sets of excitatory inputs (Schollet al., 2010). The de-novo generation of sound-offsetresponses has so far been best described for neurons inthe superior olivary complex (SOC) of the mammalian

auditory brainstem which exhibit acoustically-evoked off-set firing at the end of a sound stimulus (Grothe, 1994;Kuwada and Batra, 1999; Behrend et al., 2002; Dehmelet al., 2002; Kulesza et al., 2003; Kulesza, 2008). Theseoffsets can be generated via a post-inhibitory reboundmechanism that is initiated by strong glycinergic inputsand aided by the activation of hyperpolarization-activatedcyclic nucleotide-modulated currents (IH) and T-type cal-cium currents (Felix et al., 2011; Kopp-Scheinpflug et al.,2011). The main source of the glycinergic input triggeringthese offset responses is the medial nucleus of the trap-ezoid body (MNTB; Grothe, 1994; Kulesza et al., 2007),which on stimulation evokes IPSCs (Kopp-Scheinpfluget al., 2011) reversing at voltages around –100 mV, wellbelow the neurons’ resting membrane potential (Löhrkeet al., 2005; Yassin et al., 2014).

It has been shown in in vivo measurements that thisstrong inhibition is essential for generating short-latencyoffset responses in a substantial population of SOC neu-rons (cat: Guinan et al., 1972; bat: Grothe, 1994; rabbit:Kuwada and Batra, 1999; gerbil: Behrend et al., 2002;Dehmel et al., 2002; rat: Kulesza et al., 2003; mouse:Kopp-Scheinpflug et al., 2011; Felix et al., 2013), but theanatomic location and number of SOC neurons with offsetresponses varies across mammalian species and experi-mental approaches. In rodents, neurons with offset re-sponses are concentrated in the superior paraolivarynucleus (SPN) where current-clamp recordings in vitroshow offset firing in nearly 100% of neurons (Kopp-Scheinpflug et al., 2011; Felix et al., 2013), while SPNrecordings in vivo revealed more differential responsetypes (Behrend et al., 2002; Dehmel et al., 2002; Kopp-Scheinpflug et al., 2011; Felix et al., 2013). These discrep-ancies in observing different response types between invivo and in vitro data suggest that so far, in vitro experi-ments have missed factors that might modulate offsetresponses and generate more diverse response types.

Excitatory responses of SPN neurons in vivo have beenreported either as increased sound-evoked firing rateswithout obvious inhibition (Behrend et al., 2002; Dehmelet al., 2002) or as responses masked by inhibition, only tobe revealed after pharmacological blockade of glycinergicinputs (Kulesza et al., 2007; Jalabi et al., 2013). However,the relevance of excitatory inputs to these predominantlyoffset-responding neurons for auditory processing is un-known. Octopus cells of the ventral cochlear nucleus havebeen suggested as one source of SPN excitation basedon the contralateral origin of the sound-evoked excitation,its broad frequency tuning (Dehmel et al., 2002) and an-atomic tracing experiments (Thompson and Thompson,1991; Schofield, 1995; Saldaña et al., 2009; Felix et al.,2017). An in vitro study also reports the presence of AMPAreceptor (AMPAR)-mediated responses in the mouse SPN(Felix and Magnusson, 2016), but their function andmechanism of action are still unknown.

To gain better insight into the functional relevance ofexcitatory inputs during sound-offset encoding we per-formed immunocytochemistry, single-cell recording invivo, computational modeling, and patch-clamp record-ings in vitro. We demonstrate that the time course of slow

This work was supported by German Research Council (DFG) GrantsKO2207/3-1, SFB870-A10, and GS-82 Graduate School of Systemic Neuro-sciences GSNLMU.

*M. J. Fischl’s present address: National Institutes of Health, Section onNeuronal Circuitry, Bethesda, MD 20814.

Acknowledgements: We thank Dr. Jennifer Linden and Dr. James Sinclair forhelpful discussions and invaluable experimental support.

Correspondence should be addressed to Conny Kopp-Scheinpflug [email protected].

https://doi.org/10.1523/ENEURO.0106-19.2019Copyright © 2019 Rajaram et al.This is an open-access article distributed under the terms of the CreativeCommons Attribution 4.0 International license, which permits unrestricted use,distribution and reproduction in any medium provided that the original work isproperly attributed.

New Research 2 of 20

May/June 2019, 6(3) e0106-19.2019 eNeuro.org

mailto:[email protected]


http://creativecommons.org/licenses/by/4.0/

http://creativecommons.org/licenses/by/4.0/

NMDA receptor (NMDAR)-mediated excitation extendsinto the temporal window of post-inhibitory reboundfiring. Simultaneous activation of excitation and inhibi-tion accelerates post-inhibitory rebound responses andmakes them more tolerant against changes in soundintensity in vivo, which is a prerequisite for sound-duration tuning and level-independent gap detection(Forrest and Green, 1987).

Materials and MethodsAll experimental procedures were reviewed and ap-

proved by the Bavarian district government (TVV AZ:55.2-1-54-2532-38-13) and were done according to theEuropean Communities Council Directive (2010/63/EU).C57Bl6J mice were housed in a vivarium with a normallight dark cycle (12/12 h light/dark) and food and water adlibitum. Mice of both sexes were used for the physiologicand anatomic experiments.

ImmunohistochemistryMice (P21–P35) were anesthetized with an overdose of

pentobarbital and perfusion-fixed with 4% paraformalde-hyde (PFA) intracardially. Following overnight postfixationin 4% PFA, coronal brainstem sections including the co-chlear nucleus and the SOC of 50-�m thickness weretaken using a vibrating microtome (Leica Biosystems,VT1200S). After 3 � 10-min washes in PBS, sections weretransferred to a blocking solution containing 1% bovineserum albumin, 0.5% Triton X-100, and 0.1% saponin inPBS. For NMDAR staining, proteinase K treatment (1:1000 in PBS) for 20 min at 37°C was included beforetransferring the sections into blocking solution. Tissuewas incubated for 48 h at 4°C with primary antibodies(Table 1) diluted in blocking solution. Biocytin was labeledwith streptavidin conjugated to Cy3 (1:500 in blockingsolution). Tissue was then washed 3 � 10 min in PBS atroom temperature, before incubation for 24 h at 4°C withsecondary antibodies diluted in blocking solution. Then

sections were rinsed 3 � 10 min in PBS, and coverslippedwith Vectashield mounting medium.

Confocal microscopyConfocal optical sections were acquired with a confo-

cal laser-scanning microscope equipped with HCX PLAPO CS 20X/NA0.7 and HCX PL APO Lambda Blue 63�/NA1.4 immersion oil objectives (Leica). Fluorochromeswere visualized with excitation wavelengths of 405 nm(emission filter 410–430 nm) for amino-methylcoumarin(AMCA), 488 nm (emission filter 510–540 nm) for AlexaFluor 488, 561 nm (emission filter 565–585 nm) for Cy3,and 594 nm (emission filter 605–625 nm) for Alexa Fluor594. For each optical section, the images were collectedsequentially for the different fluorochromes. Stacks of8-bit grayscale images were obtained with axial distancesof 290 nm between optical sections and pixel sizes of120–1520 nm depending on the selected zoom factor andobjective. To improve the signal-to-noise ratio, imageswere averaged from three successive scans. RGB stacks,montages of RGB optical sections and maximum-intensity projections were assembled using the ImageJ1.37k plugins and Adobe Photoshop 8.0.1 software.

In vivo physiologyYoung adult (6–16 weeks) mice of either sex (n � 9)

were anesthetized with a subcutaneous injection of 0.01ml/g MMF (0.5 mg/kg body weight medetomidine, 5.0mg/kg body weight midazolam, and 0.05 mg/kg bodyweight fentanyl) and were placed on a temperature-controlled heating pad (WPI: ATC 1000) in a soundproofchamber (Industrial Acoustics). Depth of anesthesia wasmeasured using the toe pinch reflex and animals respond-ing were given supplemental MMF at 1/3 the initial dose.The mice were then stabilized in a custom stereotaxicdevice. An incision was made at the top of the skull, anda head post was fixed to the skull using dental cement. Acraniotomy was performed above the cerebellum to ac-

Table 1. Primary and secondary antibodies used for immunocytochemistry

Primary antibody Antigene Supplier Catalog number Host DilutionGlyT2 Synthetic from the C terminus as predicted

from the cloned rat GLYT2Millipore AB1773 Guinea pig 1:1000

GlyT2 Recombinant protein (aa1–229 of rat GlyT2) SySy 272003 Rabbit 1:1000NMDA- R2c Fusion protein from the NR2C subunit

of the NMDARR&D Systems PPS033 Rabbit 1:500

MAP2 Purified MAP2 isolated from bovine brain Acris TA336617 Chicken 1:500VGLUT1 Purified recombinant protein of rat

VGLUT 1 (aa456–560)SySy 135304 Guinea pig 1:2000

VGLUT2 Strep-Tag fusion protein of ratVGLUT 2 (aa510–582)

SySy 135402 Rabbit 1:1000

VGLUT3 Peptide (C)ELNHEAFVSPRKK, correspondingto amino acid residues 533–545of rat VGLUT3 (accession Q7TSF2);cytoplasmic, C terminus

AlomoneLabs

AGC-037 Rabbit 1:300

Secondary antibody Host species Supplier Catalog number Conjugated DilutionRabbit Donkey Dianova 711-165-152 Cy3 1:300Rabbit Donkey Dianova 711-586-152 Alexa Fluor 594 1:200Guinea pig Donkey Dianova 706-546-148 Alexa Fluor 488 1:200Guinea pig Donkey Dianova 706-166-148 Cy3 1:300Chicken Donkey Dianova 703-156-155 AMCA 1:100



cess the auditory brainstem. A ground electrode wasplaced in the muscle at the base of the neck. Glassmicroelectrodes were pulled from glass capillaries so thatthe resistance was 5–20 M� when filled with 3 M KClsolution or 2 M potassium acetate with 2.5% biocytin.Signals were amplified (AM Systems, Neuroprobe 1600),filtered (300–3000 Hz; Tucker-Davis-Technologies PC1),and recorded (�50 kHz sampling rate) with an RZ6 pro-cessor (Tucker-Davis Technologies). Python-based SPIKEsoftware (Brandon Warren, V.M. Bloedel Hearing ResearchCenter, University of Washington) was used to calibrate themulti-field magnetic speakers, generate stimuli and recordaction potentials. Stimuli consisted of pure tones (50- to100-ms duration, 5-ms rise/fall time) at varying intensity (0-to 90-dB SPL) and were presented through hollow ear barsconnected to the speakers with Tygon tubing. PSTHs wereassessed at characteristic frequency (CF) and 80-dB SPL.Spike sorting and data analysis were performed offline usingcustom MATLAB programs. At the end of the experiment,biocytin (2.5%) was deposited at the final penetration usingthe current injection mode of the amplifier (�0.5 �A, 1–2min). Thirty minutes were allowed for cellular uptake beforethe animal was perfused, and the tissue was processed forbiocytin fluorescence as described above. Recording siteswere determined using the biocytin deposition as a refer-ence for stereotaxic reconstruction.

In vitro electrophysiologyMice of either sex P15–P22 were briefly anaesthetized

with isoflurane and rapidly decapitated. Coronal brains-tem sections (150–200 �m thick) containing the SOCwere cut in an ice-cold high-sucrose, low-sodium artificialCSF (ACSF). Brainstem slices were maintained after slic-ing in normal ACSF at 37°C for 30–45 min, after whichthey were stored in a slice-maintenance chamber at roomtemperature (�22°C). Composition of the normal ACSF:125 mM NaCl, 2.5 mM KCl, 26 mM NaHCO3, 10 mMglucose, 1.25 mM NaH2PO4, 2 mM sodium pyruvate, 3mM myo-inositol, 2 mM CaCl2, 1 mM MgCl2, and 0.5 mMascorbic acid, pH 7.4, bubbled with 95% O2, 5% CO2. Forthe low-sodium ACSF CaCl2 and MgCl2 concentrationswere 0.1 and 4 mM, respectively, and NaCl was replacedby 200 mM sucrose. Experiments were conducted at 36� 1°C, maintained by an inline feedback temperaturecontroller and heated stage (Warner Instruments) with therecording chamber being continuously perfused withACSF at a rate of 1–2 ml min�1. Whole-cell patch-clamprecordings were made from visually identified SPN neu-rons using an EPC10/2HEKA amplifier (HEKA Electronik),sampling at 50 kHz and filtering between 2.9 and 10 kHz.Patch pipettes were pulled from borosilicate glass capil-laries (Warner Instruments) using a DMZ Universal elec-trode puller (Zeitz-Instuments Vertriebs GmbH), filled witha patch solution containing: 126 mM K-gluconate, 4 mMKCl, 40 mM HEPES, EGTA 5 mM HEPES, 1 mM MgCl2, 5mM Na2phosphocreatine, 0.2% biocytin, 292 mOsm, pHwas adjusted to 7.2 with KOH. For recordings of EPSCsthe internal solution contained: 135 mM Cs-gluconate,10 mM HEPES, 1 mM EGTA, 3.3 mM MgCl2, 3 mMNa2phosphocreatine, 2 mM NaATP, 20 mM TEA-Cl, 0.2%

biocytin, 300 mOsm. pH was adjusted to 7.2 with CsOH.Data were corrected for liquid junction potentials of –13.8and –13.7 mV for the potassium-based and the cesium-based internal solutions, respectively. Electrode resis-tance was between 2.4 and 6 M�. Synaptic responseswere evoked by afferent fiber stimulation with eitherconcentric or bipolar (FHC) electrodes. Voltage pulseswere generated by the HEKA amplifier and post-amplified by an isolated pulse stimulator (AM Systems).Synaptic conductances were calculated from the syn-aptic currents: G � PSC/(Vm – EPSC), with PSC beingthe postsynaptic current, Vm being the holding potential(– 60 mV for inhibition; 40 mV for excitation), EPSC beingthe reversal potential of the postsynaptic current (EE-

PSC: 0 mV; EIPSC: –100 mV).

Computational modelTwo simple, single-compartment models of a SPN neu-

ron were simulated using NEURON (version 7.5; Hinesand Carnevale, 2001). In both models, the basic set-up ofthe neuron, including the membrane properties and theionic channels, is identical to the model developed byKopp-Scheinpflug et al. (2011), available on ModelDB(https://senselab.med.yale.edu/modeldb/ShowModel.asp?model�139657), accession number 139657. In thetwo current models, the excitatory noise was removed,and four excitatory synapses were added in addition tothe inhibitory synapses already present. The synapses aremodeled using the NEURON built-in function Exp2Syn. Inboth models, the excitatory and inhibitory stimuli consistof 10 such spikes with 10-ms gaps, resulting in a totalstimulus duration of 100 ms. The excitatory and inhibitoryconductances and the respective time constants used inthe model are provided by the patch-clamp experimentsperformed for this paper. In the first model (includingAMPA currents), all four excitatory synapses are simulat-ing AMPA synapses, with a reversal potential Erev_AMPA �0 mV, a rise time constant of �1 � 0.1 ms and a decay timeconstant of �2 � 0.9 ms. In the second model (includingAMPA and NMDA currents), only two excitatory synapsesare simulating AMPA synapses, with the same parametersas in the first model; while the two other excitatory syn-apses are simulating simplified NMDA synapses, with re-versal potential Erev_NMDA � 0 mV, a rise time constant of�1 � 3 ms, and a decay time constant of �2 � 9 ms. Theconductances of the excitatory synapses were varied inboth models. Both models included 14 inhibitory syn-apses with reversal potential Erev_inh � –100 mV, a risetime constant �1 � 0.1 ms, a decay time constant �2 �2 ms and a peak conductance of 82 ns and 41 nS. Asimplified depression of the input synapses was modeledusing the steady-state depression values collected forthis paper.

Experimental design and statistical analysisIn the text data are presented in parenthesis as (me-

dian; 25/75 quartiles or as mean � SEM; test: p value)unless indicated otherwise. In the figures, data are pre-sented as medians (lines in boxes); 25/75 quartiles (box-es); and 10th/90th percentiles (whiskers) in addition toindividual data points. Statistical analyses of the data



https://senselab.med.yale.edu/modeldb/ShowModel.asp?model=139657

https://senselab.med.yale.edu/modeldb/ShowModel.asp?model=139657

were performed with SigmaStat/SigmaPlot. Normalitywas tested by the Shapiro–Wilk test. Comparisons be-tween different data sets were made depending on thedistribution of the data using parametric tests for normallydistributed data (two-tailed Student’s t test for comparingtwo groups and ANOVA for comparing three or moregroups). When the normality assumption was violated,nonparametric tests (Mann–Whitney rank-sum test forcomparing two groups and Kruskal–Wallis ANOVA onranks for comparing three or more groups) were used.Paired t tests or Wilcoxon signed rank tests were usedwhen two data sets were recorded from individual neu-rons under different conditions. Differences were consid-ered statistically significant at p � 0.05 and presented inthe figures as n.s. for nonsignificant differences and as �p� 0.05, ��p � 0.01, and ��p � 0.001 for significantdifferences. Intrinsic properties as well as postsynapticcurrent amplitudes and kinetics were analyzed usingStimfit software (Guzman et al., 2014). For data acquiredwith in vivo single-unit recording or patch-clamp record-ing; n is the number of neurons, with two to three brainslices per animal and at least three animals per group.

ResultsSPN neurons receive glycinergic and glutamatergicsynaptic input

We previously showed that only inhibitory synaptic in-put is required to generate offset responses in the SPN(Kopp-Scheinpflug et al., 2011). Here, this was corrobo-rated by immunocytochemistry showing strong expres-sion of the neuronal glycine transporter type 2 (GlyT2),which labels glycinergic synaptic terminals around thesoma of SPN neurons (Fig. 1, green boutons). However,the additional presence of excitatory synapses was con-firmed by labeling for the vesicular glutamate transporter(VGLUT) types 1–3 (Fig. 1A,D,G). While VGLUT1 andVGLUT2 positively labeled presynaptic terminals in SPN(Fig. 1B,C,E,F), the signal strength for VGLUT3 was quitelow and somatic rather than in the presynaptic boutons(Fig. 1H,I).

SPN neurons with sound-offset responses canexhibit moderate excitatory responses duringcontralateral sound stimulation

To explore the contribution of excitatory inputs to signalprocessing in neurons with sound-offset responses,spikes were recorded from single SPN neurons in anes-thetized mice in vivo, during (peristimulus) and after (post-stimulus) the presentation of sound (Fig. 2A). The sampleof 20 SPN neurons had CFs ranging from 9.25 to 50.4 kHz(Fig. 2B); 85% (17/20) of these neurons showed a burst ofincreased firing at the end of contralateral sound stimu-lation (offset responses) and little or no firing during soundpresentation (Fig. 2C,D), consistent with a prevalent in-hibitory input and the dominance of offset responsesreported in previous studies (Dehmel et al., 2002; Kuleszaet al., 2003; Felix et al., 2013). Only 3/20 SPN neurons didnot exhibit offset firing, but responded with an onset,primary-like or sustained firing pattern during sound stim-

ulation (Fig. 2E). These neurons that fired spikes onlyduring but not after the stimulus were not included infurther analyses. Of SPN neurons with offset responses,53% (9/17) additionally showed excitatory responses dur-ing sound stimulation that exceeded 5% of the respectiveneurons’ overall firing rate (Fig. 2F). These neurons will befurther referred to as ON-OFF type neurons (Fig. 2F,G,gray) and will be contrasted against neurons that exhibitoffset responses without peristimulus excitation (OFF-only type; Fig. 2F, blue). Average temporal response pat-terns show ON-OFF type neurons with either an onset ora primary-like temporal response pattern during soundfollowed by a poststimulus offset response (Fig. 2G, grayhistogram). Onset or primary-like temporal response pat-terns are associated with octopus cells in the ventralcochlear nucleus (Rhode et al., 1983) which are onesource of excitatory input to the SPN (Thompson andThompson, 1991; Schofield, 1995; Saldaña et al., 2009;Felix et al., 2017). The average response of the OFF-onlytype neurons is characterized by only little spontaneousfiring during sound followed by a slightly delayed post-stimulus offset response (Fig. 2G, blue histogram). Theratio of peristimulus to poststimulus rate is significantlylarger in the ON-OFF type neurons (ON-OFF type: 0.22;0.12/1.39; n � 9; OFF-only type: 0; 0/0.09; n � 8; Mann–Whitney rank-sum test: p � 0.002; Fig. 2H).

SPN neurons that exhibit additional excitation haveaccelerated and level-independent offset-responselatencies

The average temporal response pattern in Figure 2Gsuggests that offset-response latencies are shorter inON-OFF type neurons compared to OFF-only type neu-rons. To investigate potential differences in latency inmore detail, offset responses were recorded at CF forincreasing sound intensities (Fig. 3A,B). Indeed, offset-response latencies were significantly faster for ON-OFFtype neurons (5.55 ms, 3.80/6.97 ms, n � 8) than forOFF-only type neurons (8.84 ms, 6.38/16.75 ms, n � 9;Mann–Whitney rank-sum test: p � 0.018; Fig. 3C). How-ever, no difference was observed for the variability of theoffset-response latency (jitter) between ON-OFF andOFF-only type neurons (jitterON-OFF: 1.23 ms; 0.62/2.54ms, n � 8; jitterOFF-only: 1.11 ms; 0.45/5.23 ms; n � 9;Mann–Whitney rank-sum test: p � 0.665; Fig. 3D), sug-gesting that the intrinsic properties of the cells that regu-late precise spike firing are not the reason for thedifference in latencies.

A common feature in stimulus encoding across differentsensory modalities is that response latencies decreasewith increasing stimulus intensity (auditory: Kitzes et al.,1978; Klug et al., 2000; somatosensory: Mountcastleet al., 1957; visual: Morgan and Thompson, 1975). Incontrast, offset-response latencies of both ON-OFF type(Fig. 3A) and OFF-only type (Fig. 3B) SPN neurons did notshow a strong dependency on stimulus intensity but re-mained rather constant over a large sound intensity rangeor even showed a trend of increasing latencies with in-creasing intensities (single cell examples shown in Fig.3A,B). On average, there were no significant changes in



latency per decibel sound intensity for the ON-OFF typeneurons (–9.83 �s/dB; –19.46/5.11 �s/dB; n � 6; Fig. 3E),while changes in latencies per decibel for the OFF-onlytype neurons were significantly larger (–47.79 �s/dB;–67.37/–13.87 �s/dB; n � 6, Kruskal–Wallis ANOVA onranks followed by Dunn’s multiple comparisons versus azero change control: p � 0.032).

If the neurons firing either ON-OFF or OFF-only re-sponses represent two distinct populations, they mayinhabit different locations in the SPN. This was tested bytesting whether ON-OFF or OFF-only responding neuronshave similar or different CFs. The CFs of the offset re-sponse for ON-OFF type neurons (15.1 kHz; 12.0/21.8kHz; n � 8) were compared with those for OFF-only typeneurons (18.5 kHz; 12.3/22.3 kHz; n � 9) but no significant

difference was found (Mann–Whitney rank-sum test: p �1.000; Fig. 3F). Tonotopically organized glycinergic pro-jections from MNTB into SPN suggest that neurons tunedto high frequencies are located more medially and neu-rons tuned to low frequencies are located more laterally(Banks and Smith, 1992). As a result, neurons with differ-ent CFs should inhabit separate anatomic locations withinthe tonotopic axis of the SPN, which was not observed inthe present sample of ON-OFF and OFF-only type neu-rons and suggests, that neurons having ON-OFF or OFF-only responses are part of a continuum with the onlydifference being differently balanced excitation and inhi-bition. Response thresholds of offset responses were alsonot significantly different between ON-OFF type (32-dBSPL; 16.3/48.8-dB SPL; n � 8) and OFF-only type neu-

Figure 1. Histochemical profile of excitatory and inhibitory inputs to the SPN. A, B, D, E, G, H, Glycinergic input forms the mostprominent input to SPN (white dotted circle in A, D, G) and is depicted by neuronal GlyT2 (green) labeling. Immunolabeling formicrotubule associated protein 2 (MAP2, blue) is used as a neuronal marker in all images. Glutamatergic inputs are shown by labelingfor the VGLUTs (magenta): VGLUT1 (A–C), VGLUT2 (D–F), VGLUT3 (G–I). A–F, While both VGLUT1 (B, white arrows) and VGLUT2 (E,white arrows) positive synaptic boutons are present at the soma, VGLUT2 boutons are also seen in the neuropil (E, white arrowheads).G–I, VGLUT3 shows only weak somatic but no presynaptic bouton labeling. Scale bars � 200 �m (left images) and 20 �m (middleand right images).



rons (21-dB SPL; 18.5/28-dB SPL; n � 9; Mann–Whitneyrank-sum test: p � 0.386; Fig. 3G).

The occurrence of excitatory responses in some butnot all neurons of a nucleus with generally dominant

inhibition could also be attributed to a reduced inhibi-tory constraint in ON-OFF type neurons, resulting inhigher spontaneous firing rates compared to OFF-onlytype neurons. Inhibition in SPN neurons is provided by

Figure 2. SPN neurons with post-inhibitory rebound responses at sound offset can also show excitatory responses during sound. A,Schematic of a sound-offset encoding circuit in the mammalian brainstem. Globular bushy cell (GBC) axons project via the ventralacoustic stria (VAS) to the contralateral MNTB to form one-to-one connections via the giant calyx of Held synapses. MNTB neuronsthen project to neurons of the SPN. SPN cells also receive excitatory input from octopus cells (OCs) and possibly other, yet unknownsources (?) in the contralateral ventral cochlear nucleus via the IAS. B, Plot of CFs and auditory thresholds of all SPN neurons recordedin this sample. C–E, Raster plots and PSTHs (at CF, 80-dB SPL, 50 trials) shown for three individual SPN neurons with differentresponse patterns. C, ON-OFF type, showing a peristimulus response related to sound onset followed by a response related to soundoffset. D, OFF-only type, showing responses related to sound offset, but no peristimulus firing. E, Response type showing onlyperistimulus firing but no poststimulus response at sound offset. F, Distribution of neurons with offset responses depending on howmuch of their overall firing activity during the combined 200-ms peristimulus and poststimulus period was present within the first 20ms of the response. The dashed line represents the 5% criterion we used to classify the neurons into either OFF-only type neurons(blue: 1–9) or ON-OFF type neurons (gray: 10–17). G, Average temporal response patterns for OFF-only type neurons (blue) andON-OFF type neurons (gray). H, Peristimulus-to-poststimulus ratio for OFF-only type neurons (blue) and ON-OFF type neurons (gray).Firing rates were averaged over the whole peristimulus time window and divided by the average of the whole poststimulus timewindow. Equal firing rates in both time windows result in a ratio of zero (dotted line). Filled circles represent the ratios for the examplecells shown in C, D. Note that OFF-only neurons with ratios larger than zero exhibit spontaneous APs which also appear during theperistimulus time window but not concentrated within the first 20 ms to form an onset response. ��p � 0.01.



strong glycinergic input from MNTB neurons. Despitehigh firing rates during sound stimulation, MNTB neu-rons are spontaneously active with an average rate of20 –30 Hz (Kopp-Scheinpflug et al., 2008) which toni-cally suppresses SPN activity. Spontaneous firing ratesin SPN were not significantly different between ON-OFFtype neurons (13.25 � 3.5 Hz) compared to OFF-onlytype neurons (8.73 � 3.7 Hz; two-tailed t test: p �0.359; t � 0.946; df � 15; Fig. 3H), implying a similarstrength of inhibitory innervation across neurons.Across the population of cells tested, ON-OFF type and

OFF-only type neurons do not exhibit differences inlocation or general physiologic properties suggestingthem to belong to one cell population with differingstrength of excitatory input.

Offset-response thresholds are more sensitive thanperistimulus-response thresholds

Comparing the peristimulus and poststimulus excit-atory responses within each ON-OFF type SPN neuronrevealed that their spectral tuning largely overlapped (Fig.4A,B), which is in contrast to results from auditory cortex

Figure 3. ON-OFF type SPN neurons have shorter offset-response latencies and less level dependence than OFF-only type neurons.A, B, Raster plot at CF and changing intensity levels for representative (A) ON-OFF type and (B) OFF-only type neurons. Each dotrepresents an action potential. Gray-shaded area indicates sound duration. C, D, Distribution of (C) offset latencies and (D) jitter forOFF-only type (blue) and ON-OFF type (gray) neurons at CF/80-dB SPL. Statistical assessment in C, D was unaltered if the twoextreme values were removed. E, Average change of offset-response latency per dB intensity change for OFF-only type (blue) andON-OFF type (gray) SPN neurons. OFF-only type neurons (blue) showed a significantly larger variability of offset-response latencywith changes in intensity. Latency change/dB was not significantly different from zero (dotted line) for ON-OFF type neurons (�levelinvariant). F–H, Between SPN neurons of either OFF-only type or ON-OFF type, physiologic parameters like (F) CFs, (G) thresholds,and (H) spontaneous firing rates are not significantly different. �p � 0.05, ��p � 0.01, n.s. � non-significant.



(Scholl et al., 2010; Sollini et al., 2018) and will be dis-cussed later. There was no significant difference betweenthe peristimulus and poststimulus CFs for ON-OFF typeneurons (CFperistim: 19.8 � 4.5 kHz; n � 8; CFpoststim: 20.0� 4.9 kHz; n � 8; two-tailed paired t test: p � 0.74; t �–0.345; df � 7; Fig. 4C). Thresholds were significantly

lower for poststimulus offset responses (32 � 6-dB SPL)compared to excitatory peristimulus responses (56 �5-dB SPL; n � 8; two-tailed paired t test: p � 0.011; t �3.407; df � 7; Fig. 4D). For each ON-OFF type neuronfirst-spike latencies for the excitatory peristimulus re-sponse (4.12 � 0.2 ms) were barely faster than those of

Figure 4. Within ON-OFF type SPN neurons, peristimulus responses have similar tuning, but elevated threshold compared to theirpoststimulus offset-responses. A, B, Frequency-intensity response maps for an individual ON-OFF type SPN neuron plotted for theperistimulus response (A) and for the poststimulus response after sound cessation (B). C, Linear correlation between peristimulus andpoststimulus CFs for individual ON-OFF type SPN neurons. D, Thresholds for the responses during sound (peristimulus, white circles)are higher than thresholds for responses after sound termination (poststimulus, gray circles). E, F, Latencies and jitter were notsignificantly different between peristimulus and poststimulus response of individual ON-OFF type SPN neurons. �p � 0.05, n.s. �non-significant.



the offset response following sound cessation (5.64 � 1.0ms; two-tailed paired t test: p � 0.188; t � 1.458; df � 7;Fig. 4E), indicating a high speed in generating offsetresponses despite the additional synaptic delay arisingduring the sign conversion in the MNTB. Jitter as a mea-sure of the temporal precision of the first spike in theresponse was also not significantly different between theperistimulus (0.64 � 0.17 ms) and the poststimulus (1.45� 0.35 ms) response (two-tailed paired t test: p � 0.111;t � 1.822; df � 7; Fig. 4F). The mean latency of �4 ms forthe peristimulus response is similar to that of other SOCneurons in mouse (Kopp-Scheinpflug et al., 2008). Al-though the average temporal response pattern of theON-OFF type neurons shows a primary-like pattern (Fig.2G), 5/8 ON-OFF type neurons show onset firing patterns.To investigate whether primary-like or onset-responsescan influence the offset response, in vivo recordings ofspikes do not provide sufficient information about possi-ble subthreshold activity that lasts throughout the stimu-lation, which will more appropriately be measured bywhole-cell patch-clamp recordings in vitro.

Strength of inhibition outweighs excitation in SPNneurons

Whole-cell patch-clamp recordings in vitro were madeof SPN neurons. Neurons had resting membrane poten-tials of –61.74 � 0.79 mV, an average input resistance of

74.39 � 10.94 M� and an average membrane capaci-tance of 65.63 � 7.78 pF (n � 23). Glutamatergic EPSCswere regularly activated in neurons throughout the SPN.For electrical stimulation, a concentric stimulating elec-trode was placed on the intermediate acoustic stria (IAS)medial to the SPN and just dorsal to the MNTB (Fig. 5A).EPSCs were pharmacologically isolated by adding theGABAA receptor blocker SR95531 (20 �M) and the glycinereceptor blocker strychnine (1 �M) to the bath solution(Fig. 5B). Glycinergic IPSCs were evoked by directelectrical stimulation of the ipsilateral MNTB using a con-centric stimulating electrode and were pharmacologicallyisolated by adding the AMPAR blocker 6,7-dinitro-quinoxaline-2,3-dione (DNQX; 10 �M) and the NMDARblocker D-2-amino-5-phosphonopentanoic acid (D-AP5;50 �M) to the bath solution (Fig. 5C). At physiologicholding voltages near the neurons resting membrane po-tential (–60 mV), IPSC amplitudes were significantly largerthan EPSCs (IPSC: 2.46; 0.98/3.77; n � 12; EPSC: 0.23;0.20/0.65; n � 11; Mann–Whitney rank-sum test: p �0.001; Fig. 5D). The considerable difference between thestrength of excitation and that of inhibition was even moreobvious when both were expressed as conductances(IPSG: 45.65; 18.22/70.02; n � 12; EPSG: 4.26; 3.72/12.08; n � 11; Mann–Whitney rank-sum test: p � 0.001;Fig. 5E). The decay time constants of IPSCs were signif-icantly slower than that of EPSCs (TauIPSC: 1.09; 0.90/

Figure 5. Comparison of synaptic strength between EPSCs and IPSCs in SPN neurons. A, Schematic of the sound-offset encodingcircuit depicting the positions of the stimulating electrodes for eliciting either excitation or inhibition. B, C, Voltage-clamp traces ofpharmacologically isolated EPSCs (B) evoked by stimulating IAS and IPSCs (C) evoked by stimulating MNTB (average of 10 traces).Black traces indicate the blockade of (B) EPSCs or (C) IPSCs. D, Average IPSC (blue) and EPSC (red) amplitudes measured in responseto maximum stimulation. E, EPSCs and IPSCs expressed as conductance reveals that mean IPSG values are more than five times largerthan EPSGs. F, Average decay time constants (�) of EPSCs and IPSCs. TauIPSC is significantly slower than TauEPSC. ��p � 0.001.



1.21; n � 12; TauEPSC: 0.49; 0.36/1.24; n � 11; Mann–Whitney rank-sum test: p � 0.038; Fig. 5F).

NMDAR-mediated currents are present in SPNneurons

Glutamate released from excitatory synaptic inputs typ-ically activates AMPAR and/or NMDAR with fast and slowkinetics, respectively. Current-clamp recordings of SPNneurons near the neuronal resting potential suggestedthat excitatory responses are primarily mediated byAMPARs (Felix and Magnusson, 2016). Here, we usedimmunocytochemistry to probe for the presence of

NMDARs and performed voltage-clamp recordings to as-sess the strength of NMDAR-mediated currents at differ-ent membrane potentials. NMDAR expression is evidentin mature SPN neurons, although weaker compared toneurons in the MNTB or LSO (Fig. 6A–D). Electrophysio-logical measurement of NMDA currents in whole-cellvoltage-clamp mode revealed the characteristic voltagedependence causing larger currents once the membranevoltage reaches depolarized values (Fig. 6E,F) andshowed that they were sensitive to the NMDAR antagonistD-AP5 (Fig. 6E). NMDA currents in SPN neurons (35.0pA;25.2/60.3pA; n � 14; Fig. 6G) were smaller than in MNTB

Figure 6. SPN neurons express NMDARs which mediate moderate EPSCs at depolarized voltages. A, Low-power image of SOCshowing NMDA immunoreactivity (magenta), which was strong in the MNTB and the lateral superior olive (LSO), and to a lesser degreepresent in the SPN. GlyT2 labeling (green) identifies the outline of the SPN. MAP2 (blue) was used as a neuronal marker. B–D, Highermagnification images show that NMDARs are present in SPN neurons. E, Voltage-clamp traces of pharmacologically isolated NMDAcurrents at –60 mV (gray) and at �40 mV (black). Currents were blocked by D-AP5 (green). F, Average NMDA currents show thetypical nonlinearity due to the Mg2� block at hyperpolarized membrane voltages. G, Amplitude of NMDA currents at �40 mV. H,NMDA conductance. I, Decay time constants (�) of NMDA currents at �40 mV. Scale bars � 200 �m (A) and 20 �m (B–D).



(Steinert et al., 2010), but similar to LSO (Alamilla andGillespie, 2011; Pilati et al., 2016) and MSO (Smith et al.,2000; Couchman et al., 2012). NMDA conductances cal-culated from these currents ranged from 0.36 to 1.88 nS(0.65; 0.47/1.12 nS; n � 14). The small, yet prevalentNMDA currents resemble the general decline of NMDAcurrents in the auditory brainstem following hearing onset.However, as Steinert and colleagues have shown for neu-rons in the neighboring MNTB, after the initial reduction inamplitudes, NMDA currents reach a steady state with nosigns of further decline after about two weeks of age(Steinert et al., 2010), corroborating the presence ofNMDA current in the SPN for ages of 15 d and older.Decay time constants of the NMDAR-mediated EPSCsranged from 7.34 to 13.69 ms (9.21 ms; 7.74/12.79 ms; n� 8; Fig. 5H). In later experiments AMPAR and NMDARresponses will be activated and blocked in unison by acocktail of DNQX/D-AP5 and together will be contrastedagainst glycinergic inhibition to reveal the contribution ofexcitation to ON-OFF type SPN neurons.

Differential short-term plasticity between excitatoryand inhibitory SPN synapses

To test for a balance between excitatory and inhibitoryinputs in an adapted, more physiologic state, pharmaco-logically or electrically isolated inhibitory, excitatory andNMDAR-mediated synaptic currents were evoked in SPNneurons by applying 50-pulse fiber stimulation trains at100 Hz (Fig. 7A). Since NMDAR-mediated currents canonly be activated at depolarizing potentials positive to theEPSC reversal potential, they are shown as outward cur-rents in the green trace in Figure 7A. For better visualiza-tion NMDA currents are flipped in the superimposedenlargement of all three current types for the first 100 ms ofthe response (Fig. 7B). Both fast inhibitory and excitatorycurrents showed pronounced short-term depression, calcu-lated from a train of IPSCs and EPSCs, respectively, andnormalized to the first peak (Fig. 7C,D). In contrast NMDAR-mediated currents showed very little depression (Fig. 7C,D).The time constant of depression was fastest for EPSCs (1.92ms; 1.41/2.39 ms; n � 9), a little slower in IPSCs (2.88 ms;2.53/3.23 ms; n � 17) and very slow in NMDAR-mediatedcurrents (10.31 ms; 4.01/16.45 ms; n � 5; Kruskal–WallisANOVA on ranks followed by Dunn’s multiple comparisons;p � 0.001; Fig. 7E). The average level of steady state de-pression was similar between EPSCs and IPCSs, but signif-icantly lower in in NMDAR-mediated currents (inhibition:66.70%; 62.57/72.05%; n � 17; excitation: 74.03%; 71.54/81.42%; n � 9; NMDA: 42.36%; 25.87/55.40%; n � 5;Kruskal–Wallis ANOVA on ranks followed by Dunn’s multiplecomparisons; p � 0.001; Fig. 7F).

Across the SPN, IPSCs were always larger than EPSCs.The position of each recorded neuron within the SPN waslogged by taking a picture of the electrode position in 4�magnification. The SPN was then divided into 4 quad-rants, which were correlated to the size of the currents. Nocorrelation was found, suggesting a homogeneous distri-bution of excitatory inputs across the SPN with no dis-tinct, spatially segregated subpopulations. The moderate,excitatory inputs to SPN neurons become visible as pat-

terns of increased firing activity during sound presentationin about half of the offset-responding SPN neurons. Thefact that these neurons have peristimulus as well as post-stimulus responses raises new questions: do glutamater-gic inputs, that drive peristimulus excitatory responses,interact with peristimulus inhibitory inputs that drive thepoststimulus offset response, and if so, what is theirimpact on the offset response? To specifically addressthe role of NMDAR-mediated currents, compared toAMPAR-mediated currents, we fed the data acquired byvoltage-clamp recording into a computational model tosimulate both types of excitatory synapses and the inhib-itory inputs to an SPN neuron.

Computational modeling suggests that moderateand slow excitation affects offset-response latency

A basic Hodgkin–Huxley model of SPN neuron firingwas developed to test the effect of excitatory inputs ofvariable strength that are present in addition to inhibitoryinputs. All synaptic conductances used for the followingsimulations were taken from the results presented in Fig-ures 5–7. Conductance evoked by a single pulse, thesynaptic depression during trains of synaptic stimulationand the duration of the stimulus train were taken intoconsideration while determining the range of conduc-tances to be used in the model. Inhibitory conductances,recorded in vitro after a single pulse ranged from 11.5 to80.5 nS with a median of 45 nS. We estimated the syn-aptic depression during a 100-Hz stimulation for 100 msto be around 50%, resulting in a conductance of 41 nS(50% of 80.5 nS). For longer stimulation of 500 ms, asteady state depression estimated at 64%, results inminimal possible conductance values of 3.9 nS (64% of11.5 nS). This is approximately the value where the inhi-bition becomes strong enough to generate a reboundspike in the model depicted in Figure 8E. Excitatory con-ductances were subjected to a similar approach, withminimum experimentally acquired excitatory conduc-tances of 0.4 nS (minimal measured value with maximumdepression) to 12.6 nS (maximal measured value with nodepression). For NMDA currents, the range between theminimal measured value with maximum depression andmaximum measured value with no depression was 0.3–1.9 nS. However, since NMDA currents depress only littleand even initially facilitate (Fig. 7C,D), an average of 11%facilitation was added to the maximum measured value,providing an NMDA conductance of 2.1 nS. The valuesfrom in vitro experiments might be underestimating thesynaptic conductances, as axons will be cut during theslicing procedure. To account for this caveat, we takeadvantage of the model to simulate a broader range ofconductances. These physiologically feasible core con-ductances for inhibition, excitation and NMDA are cov-ered in the matrices in Figure 8E,F. When using aninhibition-only model (Kopp-Scheinpflug et al., 2011) withadapting inputs, SPN neurons reliably fired a burst ofoffset responses following a 100-ms train of IPSCs pre-sented at 100 Hz (Fig. 8A). The original model wasamended in the full model by adding adapting AMPAR-mediated and NMDAR-mediated currents (Fig. 8B). This



model, now including inhibition and excitation, causedneurons to fire spikes not only at the end of the stimulustrain but also at the onset of stimulation (Fig. 8B). Com-paring the offset responses elicited by either model re-vealed that the model incorporating excitation (Fig. 8C,full model) generated shorter offset-response latencies.

Adding stochastic noise to either model introduced a jitterto the offset-response latencies (Fig. 8D). However, theaverage latencies for the full model (7.55 � 0.36 ms) werestill shorter compared to the inhibition only model (9.12 �0.33 ms) by 1.57 ms (two-tailed t test: p � 0.0012; t �3.245; DF: 998; Fig. 8D, red and blue vertical lines).

Figure 7. Short-term plasticity of excitatory and inhibitory inputs to SPN neurons. A, Examples of inhibitory (blue), excitatory (red), andNMDAR (green)-mediated responses to 50 stimulations at 100 Hz (averages of 10 repetitions). Gray trace depicts the lacking NMDARmediated response at –60 mV. B, Overlaid and magnified first responses of the examples shown in A. The NMDA trace (green) has beenflipped to symbolize its excitatory nature. C, Normalized and averaged current amplitudes to each of the 50 pulses of the 100-Hz train forNMDA currents (green), EPSCs (red), and IPSCs (blue). D, First 100 ms of the plots shown in C. E, Time constant (�) of the rate of depressionacquired from fitting exponential decay functions to the functions shown in C. F, Summary of synaptic currents steady-state depression: 100%– (average current amplitudes in response to the last five pulses � 100). �p � 0.05, ��p � 0.01, ��p � 0.001, n.s. � non-significant.



To assess how the balance between excitation andinhibition affects offset-response latencies, both conduc-tances were independently varied in strength and thecorresponding changes in latencies shown in a heat map(Fig. 8E; model including AMPA currents). Specific com-binations elicited both onset and offset responses andwere thus most comparable to the ON-OFF type neuronalresponses in vivo, while in most cases no onset responsewas generated. Comparing the offset-response latenciesfor vertical columns of a specific inhibitory conductance(e.g., 41 nS in Fig. 8E) and increasing AMPA conduc-tances revealed that varying the AMPA conductancealone mediates the spike at stimulus onset, but did notsignificantly shorten the latencies of spikes in the offset-response (Fig. 8E). On the contrary, increasing the AMPAresponse beyond the physiologic estimates will prolongthe offset-response latencies due to the occurrence ofperistimulus spikes that interfere with the refractory pe-

riod of the offset-response spikes. The shortest offset-response latencies were obtained with a combination ofstrong inhibition (41 nS) coupled with both AMPA andNMDA conductances (Fig. 8F). In the latter condition,varying the AMPA conductance between 0 and 2.56 nS(y-axis) did not change the offset-response latency. How-ever, increasing the NMDA conductance (x-axis) for any ofthese AMPA conductances increasingly shortened theoffset-response latency. Increasing the AMPA conduc-tance further (�5.1 nS), caused continuous peristimulusspike firing which then prolonged offset-response laten-cies due to refractory interactions. A combination of inhi-bition and NMDA conductance alone did not have aneffect on offset-response latencies (data not shown).

These simulations suggest that although offset re-sponses can be generated by exclusively activating thechloride conductance, the slow NMDA conductance thataccompanies fast AMPA-mediated responses serve to

Figure 8. Excitatory input modulates offset-response timing in a computational model of SPN neurons. A, Voltage trace of a SPN neuron’sresponse to a train of 10 stimuli using an inhibition only model shows 10 IPSPs followed by a burst of spikes at stimulus offset. B, Voltagetrace of a SPN neuron’s response to a train of 10 stimuli using the amended “full” model including AMPAR-mediated and NMDAR-mediatedexcitation shows a spike at stimulus onset followed by 10 IPSPs and then by a burst of spikes at stimulus offset. C, Larger temporalresolution of the first two action potentials of the offset response (gray shaded area in A, B). D, Distributions of offset-response latenciesfor the inhibition only model (blue) and the full model (red) for 500 repetitions when stochastic noise was added. The red and blue solid linesrepresent the mean offset latencies for the full and inhibition only model, respectively. E, Heat map showing changes in offset-responselatencies in relation to the strength of inhibition (x-axis) and AMPAR-mediated excitation (y-axis). The first number for each stimuluscombination depicts the number of peristimulus spikes followed by the number of poststimulus spikes in the offset response. Darker greenstands for longer latencies and lighter green for shorter latencies. F, Heat map showing changes in offset-response latencies in relation tothe strength of AMPAR-mediated responses (y-axis) and NMDAR-mediated responses (x-axis). Numbers and color code are the same asin E. White areas in E specify stimulus combinations that do not generate offset responses.



accelerate the offset response. Overall, a longer EPSCdecay time results in a shorter latency, as the effect of theexcitation is carried forward into the rebound, providingadditional excitatory drive. We found that the shortestoffset-response latency was generated with simulatedEPSC decay time constants of 30 ms, but even EPSCdecay times of 10 ms accelerate the offset response.

Simultaneous activation of excitation and inhibitionin vitro results in a reduced net hyperpolarization ofPSPs and shorter offset-response latencies

Excitatory inputs to SPN neurons are not a prerequisitefor generating offset responses (Kopp-Scheinpflug et al.,2011). However, computational modeling (Fig. 8) sug-

gests that excitation might shape the temporal precisionof the offset responses. Here we compared synapticallyevoked offset responses in SPN neurons in control con-dition and during blockade of excitatory inputs by simul-taneously stimulating both the MNTB and IAS using afork-like bipolar stimulating electrode (Fig. 9A). At the endof a 100 ms, 100-Hz train of stimuli, SPN neurons reliablygenerated an offset response consisting of a burst ofspikes (Fig. 9B). The blockade of glutamatergic transmis-sion during 100-Hz stimulation for 100 ms (Fig. 9B, redtrace) caused a hyperpolarizing drop in the net amplitudeof the evoked PSPs (Fig. 9B). In control condition, thestimulation of synaptic input triggered a combined re-sponse of excitation and inhibition resulting in a net hy-

Figure 9. Excitation improves offset-response timing. A, Schematic sound-offset circuit depicting the position of the fork-electrode forsimultaneously stimulating excitation and inhibition. B, 100-Hz synaptic stimulation (black arrows) elicited an offset response in control(black) and during blockade of excitation (red). Stimulus artifacts are removed for clarity. Postsynaptic potentials were more depolarized incontrol (black). C, Average difference in amplitudes of PSPs (not spikes) between blockade of excitation and control was plotted againststimulus number within each train (circles are averages of 10 trials/cell; n � 5 cells). Black line and gray-shaded area represent mean �SEM. D, E, Higher temporal resolution of the offset responses of the cell shown before (D) and after (E) the blockade of excitation. F,Reliability of offset responses in 10 consecutive traces (as shown in D, E; 100%: at least one rebound spike/trace); n � 9. G, Averagenumber of rebound spikes (10 trials; n � 9). H, Resting membrane potential averaged over 10 ms before the start of synaptic stimulations(n � 9). I, Normalized increase in offset-response latencies after blocking excitation (n � 6). �p � 0.05, n.s. � non-significant.



perpolarization of –80.9 � 0.3 mV (n � 5) averaged overthe stimulus train. However, blocking excitation causedthe PSPs to drop by 2.6 � 0.1 mV (n � 5) toward morehyperpolarizing voltages (Fig. 9B,C). The voltage differ-ence between drug and control condition revealed asmall, yet sustained depolarizing drive over the wholestimulus time (Fig. 9C), consistent with our simulationswhich predicted the presence of a small, slow excitatoryconductance which extends into the temporal window ofthe offset response.

Blockade of excitatory inputs (only the last IPSPs of thetrains shown in Fig. 9D,E) did not significantly change thereliability of generating an offset response: in 10 consec-utive input trains, offset bursts were generated in 90 �9% of trains in controls and 86 � 7% of trains duringblockade of excitation (Mann–Whitney rank-sum test: p �0.551; Fig. 9F). The number of spikes within each burst ofthe offset response was also not significantly differentbetween the control condition (2.30; 1.7/4.6; n � 9) andfollowing the blockade of excitation (1.75; 1.0/3.8; n � 8;Mann–Whitney rank-sum test: p � 0.359; Fig. 9G). Toprobe whether tonically active excitation is present in theSPN, we compared the resting membrane potential inabsence of synaptic stimulation before and during block-ade of excitation, but no significant difference was found(Vrestcontrol: –61.8 � 2.4 mV; Vrestdrug: –61.1 � 2.0 mV;two-tailed t test: p � 0.832; t � 0.215; df � 15; Fig. 9H).Functionally, the blockade of excitation caused longeroffset-response latencies. Depending on the exact posi-tioning of the stimulating electrode and patch pipette ineach in vitro preparation, the latencies of synaptically-evoked offset-responses varied between cells from 5.6 to68.5 ms (control) and from 7.2 to 84.2 ms (DNQX/AP5).However, each paired recording showed an increase inlatency during blockade of excitation and resulted in asignificant increase of 30.68 � 6.71% in offset-responselatency when excitation was blocked (two-tailed paired ttest: p � 0.01; t � –4.571; df � 4; Fig. 9I).

In conclusion, our data show that the right balance ofmoderate, slow excitation and strong inhibition will accel-erate acoustically-evoked offset responses. Besides thisfaster offset firing, the presence of additional excitationalso significantly reduces intensity-dependent changes inresponse latency. Based on the distributions of synapticinputs within the SPN, similar CFs, thresholds and spon-taneous rates between neurons with ON-OFF type andOFF-only type responses, we conclude that the differentresponse patterns are not arising from two different typesof neurons but rather reflect differences in the balance ofexcitation and inhibition.

DiscussionSPN neurons are reliable detectors of sound offsets

and respond with a burst of spikes time-locked to the endof the stimulus. The latencies of these offset responsesare short and level-invariant over a large range of supra-threshold sound intensities, which are prerequisites forsound-duration encoding and gap-detection (Forrest andGreen, 1987; Brand et al., 2000; Faure et al., 2003; Pérez-González et al., 2006). For half of the neurons with offset

responses, additional excitatory inputs were observed inin vivo recordings. These neurons exhibited even shorteroffset-response latencies and stronger level invariance ofless than 10 �s/dB. Pharmacological manipulation andcomputational modeling showed that while inhibitionalone can reliably trigger a post-inhibitory rebound re-sponse at stimulus offset; excitation alone will not gener-ate an offset response. However, the presence of slow,NMDAR-mediated excitation facilitates the rebound de-polarization and speeds up the latencies of offset re-sponses.

Why don’t all SPN offset cells show peristimulusexcitation?

Shorter latencies and greater resistance to level depen-dent latency shifts due to additional excitation provideadvantages for computation of sound duration and fordetection of silent gaps in noise. So why do not all SPNneurons with offset responses benefit from this advan-tage? Our immunocytochemical data (Fig. 1) show uni-form distributions of excitatory and inhibitory inputsthroughout the SPN, suggesting that the observed differ-ences in the strength of excitation and inhibition betweenON-OFF type and OFF-only type neurons may not mani-fest in different synaptic input patterns. This is in agree-ment with the lack of differences in CFs, spontaneousrates or thresholds between ON-OFF type and OFF-onlytype neurons. Whether or not neurons in the SPN could beclassified into distinct subpopulation has been discussedfor 40 years. Ollo and Schwartz (1979) used Golgi impreg-nations to assess the morphology of mouse SPN neurons.They described slight morphologic differences such astriangular, elongated or polygonal shapes, but stated thatthese were “not sufficiently distinct to warrant division intodifferent cell types.” More complex approaches combinedthe morphologic description with either immunostainingfor the neurotransmitter used by SPN neurons (Helfertet al., 1989) or with neural tracing experiments (Schofield,1991). As a result five SPN cell types that project to theipsilateral inferior colliculus (IC) were described: (1) largeround glycinergic neurons, (2) large round GABAergicneurons, (3) small, round, projecting bilaterally to ICglycine-negative neurons, (4) small neurons with only ip-silateral IC projections, (5) small neurons with ipsilateral ICand contralateral cochlear nucleus projections (Helfertet al., 1989; Schofield, 1991). Another approach to clas-sify SPN neurons was taken by Felix and co-authors, whoreported only subtle differences in the intrinsic propertiesof SPN neurons and suggested that these might becaused by gradients of potassium currents (Felix et al.,2013). Whether or not the bursting cells in the dorsolateralSPN region can form a particular subtype or are stillsubject to developmental change is not yet clear. In con-clusion, the occurrence of ON-OFF type responses in asubset of SPN offset cells could be the result of differ-ences in the balance of existing excitatory and inhibitoryinputs. Previous studies have shown that the assessmentof the excitatory-inhibitory balance might be confoundedthrough the use of anesthetics which either block NMDAcurrents (ketamine-based anesthesia) or alter inhibition



(barbiturate based anesthesia) (Steinert et al., 2008; Felixet al., 2012). Here, we used a fentanyl-based anesthesia,which binds �-opioid receptors and therefore should notdirectly interfere with either excitation or inhibition in theauditory brainstem.

We conclude that the occurrence of ON-OFF type andOFF-only type neurons in SPN is neither a developmentalnor an experimental (anesthesia-based) effect. Instead,ON-OFF or OFF-only responses in SPN neurons are likelycaused by differences in the strength of excitation andinhibition whose activity- or context-dependent controlwill have to be investigated to further our knowledge onencoding sound offsets.

Facilitating the post-inhibitory reboundThe excitatory input alone does not generate offset

action potentials, yet it is sufficient to modify the offsetresponses generated via a post-inhibitory rebound mech-anism (Felix et al., 2011; Kopp-Scheinpflug et al., 2011). Asubthreshold inhibitory input aiding a subthreshold excit-atory input to increase temporal precision is generallyreferred to as post-inhibitory facilitation and has beenshown in other SOC neuron types (Dodla et al., 2006;Beiderbeck et al., 2018). This is not the case in ourpresent SPN data, since here the inhibitory input on itsown is sufficient to generate offset spikes. However, theslower time course of the additional excitation mediatedvia NMDARs allows it to extend its depolarization into thetemporal window of the post-inhibitory rebound justenough to accelerate the offset response. Such an exten-sion of excitation into the post-inhibitory rebound wouldrequire sustained excitatory responses throughout theduration of the stimulus. The fact that the peristimulusexcitatory responses observed in the ON-OFF type neu-rons in vivo occur as either sustained (38%) or onset(62%) responses is likely due to most of the excitatoryinputs being subthreshold caused by dominant inhibitionduring sound presentation, rendering subthreshold EP-SPs invisible to our single cell extracellular assessment ofspiking. Such subthreshold, peristimulus excitation wasdemonstrated by pharmacological blockade of inhibitionin vivo (Kulesza et al., 2007). Based on these previousfindings and our present results, it seems likely that, atleast part of the peristimulus excitatory inputs are sub-threshold and may serve modulatory functions rather thanto form a reliable representation of sound onset. Thiswould be in agreement with the suggestion that soundonsets and offsets are encoded in segregated pathwayswithin the auditory brain (Scholl et al., 2010; Anderson andLinden, 2016; Kopp-Scheinpflug et al., 2018; Sollini et al.,2018). While the SPN strongly qualifies for encoding off-sets, sound onset information is likely provided by differ-ent neuronal pathways such as for example the ventralnucleus of the lateral lemniscus.

The behaviorally relevant readout in the present dataset is the reduction in offset-response latencies, whichcan decrease gap-detection thresholds (Yassin et al.,2014). The slow, lasting excitation responsible for thefaster offsets, could represent the underlying cellularequivalent to adding background noise in a behavioral

gap detection task, where gap detection thresholds havebeen significantly decreased (better), compared with acondition when no background noise was provided (Hor-witz et al., 2011).

However, faster is not always better, especially whentemporal precision (as measured by jitter) is equally good.An alternative interpretation is that the modulation ofoffset-response latencies might present a tool to acceler-ate or delay the time point of the offset response depend-ing on the balance between excitation and inhibition.Such a shift in latency could present a homeostatic ad-aptation to a changing balance between excitation andinhibition as might occur during aging or following acous-tic trauma.

Reducing the level dependency of the offsetresponse

The offset-response latency depends on sufficient hy-perpolarization and acceleration of the membrane timeconstant via recruiting additional ionic conductancessuch as IH (Kopp-Scheinpflug et al., 2011). Hyperpolariza-tion of the membrane activates IH, which accelerates ofthe membrane time constant and shortens the offset-response latency. If, however, IH is already maximallyactivated by hyperpolarization of the membrane, any fur-ther hyperpolarization might prolong the offset-responselatency as it takes the membrane longer to move from–100 mV (SPN IPSP reversal potential) to � 40 mV (so-dium channel activation voltage) than for example from–80 to �40 mV. This is likely what happens with the trendof increasing offset-response latencies with increasingintensity seen in Figure 3A,B. At higher sound intensities,multiple MNTB axons will be recruited and due to inputsummation, the net-depression of inhibitory inputs in SPNneurons will be reduced, resulting in strong hyperpolar-ization of the membrane voltage even at the end of thestimulus train. The recruitment of additional excitatoryconductances as observed in the present study only athigher stimulus intensities is likely to have a similar effectas IH, in shunting the inhibition, providing a depolarizingdrive and shortening the offset-response latencies.

In contrast, at lower sound intensities only few MNTBaxons might be recruited to inhibit SPN neurons. Theseinhibitory inputs will depress over time as shown for theSPN in this study or for the LSO and MSO in other studies(Couchman et al., 2010; Walcher et al., 2011; Robertset al., 2014) and an offset response is generated at thetime point when the intrinsic depolarizing drive dominatesthe hyperpolarization. In the experimental in vitro condi-tion, especially for long stimulations of several hundredmilliseconds, this can lead to rebound responses evenbefore the end of the stimulus train (unpublished obser-vations). In vivo, such rebound response before the end ofa sound was never observed, suggesting that in vivo thedepression of collective inhibitory inputs is low and thehyperpolarization at the stimulus end is strong.

Implications of level-invariant offset responses forgap-detection and sound-duration encoding

The threshold for detecting brief silent gaps in noiseprovides a valuable analytical tool to measure temporal



resolution in auditory processing (Moore, 1997). Gaps asshort as 2–3 ms can be detected by humans (Penner,1977) and rodents (Ison, 1982) and it has been suggestedthat to detect a 3-ms gap, auditory neurons need toencode onsets and offsets of sounds with a temporalacuity of �1 ms (Oertel et al., 2017). Gap-in-noise stimulihave also proven very helpful in determining the ability ofthe auditory system to encode sound offsets as a param-eter independent of sound onsets (Pratt et al., 2005).

Behaviorally, sound offsets are an important cue forsound-duration encoding. Neurons in the auditory mid-brain that are sensitive to sound duration act as coinci-dence detectors that only fire action potentials ifexcitatory postsynaptic responses evoked by the onset ofsound temporally coincide with excitatory postsynapticresponses evoked by the offset of sound (Casseday et al.,1994; Aubie et al., 2009, 2012, 2014; Sayegh et al., 2011;). Duration sensitive neurons have been classified accord-ing to their ability to preferably encode sounds of differentdurations and are referred to as short-pass, bandpassand long-pass duration tuned neurons. According to thecurrent models of sound-duration encoding, offset exci-tation is needed in the coincidence detection bandpassmechanism but not in the anti-coincidence detectionshort-pass mechanism (Aubie et al., 2012). This is inagreement with previous results showing that temporallyprecise SPN offset responses vary with stimulus durationand provide an inhibitory projection to the auditory mid-brain (Kadner et al., 2006; Kopp-Scheinpflug et al., 2011)where they are suggested to generate a post-inhibitoryrebound excitation (Pollak et al., 2011). Interestingly, boththe detection of gaps (Moore, 1997) and the discrimina-tion of different sound durations (Klink and Klump, 2004)are relatively independent of changes in suprathresholdsound level. Especially the duration discrimination of lon-ger sounds (50 ms) has been reported to not show anintensity effect (Henry, 1948). The level-independenceand short latency of SPN offset responses over an ex-tremely large range of intensities as shown in the presentstudy provide a perfect function for the ON-OFF type SPNneurons in the encoding of longer sound durations. Defi-cits in SPN offset encoding might therefore result in diffi-culties in processing sound offsets in downstreamauditory areas like the IC or the MGB (Anderson andLinden, 2016).

ReferencesAlamilla J, Gillespie DC (2011) Glutamatergic inputs and glutamate-

releasing immature inhibitory inputs activate a shared postsynap-tic receptor population in lateral superior olive. Neuroscience 196:285–296.

Anderson LA, Linden JF (2016) Mind the gap: two dissociable mech-anisms of temporal processing in the auditory system. J Neurosci36:1977–1995.

Aubie B, Becker S, Faure PA (2009) Computational models of milli-second level duration tuning in neural circuits. J Neurosci 29:9255–9270.

Aubie B, Sayegh R, Faure PA (2012) Duration tuning across verte-brates. J Neurosci 32:6373–6390.

Aubie B, Sayegh R, Fremouw T, Covey E, Faure PA (2014) Decodingstimulus duration from neural responses in the auditory midbrain.J Neurophysiol 112:2432–2445.

Banks MI, Smith PH (1992) Intracellular recordings from neurobiotin-labeled cells in brain slices of the rat medial nucleus of the trap-ezoid body. J Neurosci 12:2819–2837.

Behrend O, Brand A, Kapfer C, Grothe B (2002) Auditory responseproperties in the superior paraolivary nucleus of the gerbil. JNeurophysiol 87:2915–2928.

Beiderbeck B, Myoga MH, Müller NIC, Callan AR, Friauf E, Grothe B,Pecka M (2018) Precisely timed inhibition facilitates action poten-tial firing for spatial coding in the auditory brainstem. Nat Commun9:1771.

Brand A, Urban R, Grothe B (2000) Duration tuning in the mouseauditory midbrain. J Neurophysiol 84:1790–1799.

Casseday JH, Ehrlich D, Covey E (1994) Neural tuning for soundduration: role of inhibitory mechanisms in the inferior colliculus.Science 264:847–850.

Cavaco S, Lewicki MS (2007) Statistical modeling of intrinsic struc-tures in impacts sounds. J Acoust Soc Am 121:3558–3568.

Couchman K, Grothe B, Felmy F (2010) Medial superior olivaryneurons receive surprisingly few excitatory and inhibitory inputswith balanced strength and short-term dynamics. J Neurosci 30:17111–17121.

Couchman K, Grothe B, Felmy F (2012) Functional localization ofneurotransmitter receptors and synaptic inputs to mature neuronsof the medial superior olive. J Neurophysiol 107:1186–1198.

Dehmel S, Kopp-Scheinpflug C, Dörrscheidt GJ, Rübsamen R (2002)Electrophysiological characterization of the superior paraolivarynucleus in the Mongolian gerbil. Hear Res 172:18–36.

Deneux T, Kempf A, Daret A, Ponsot E, Bathellier B (2016) Temporalasymmetries in auditory coding and perception reflect multi-layered nonlinearities. Nat Commun 7:12682.

Denève S, Machens CK (2016) Efficient codes and balanced net-works. Nat Neurosci 19:375–382.

Dodla R, Svirskis G, Rinzel J (2006) Well-timed, brief inhibition canpromote spiking: postinhibitory facilitation. J Neurophysiol 95:2664–2677.

Eggermont JJ (2015) Auditory temporal processing and its disorders.New York: Oxford University Press.

Elgueda D, Duque D, Radtke-Schuller S, Yin P, David SV, ShammaSA, Fritz JB (2019) State-dependent encoding of sound and be-havioral meaning in a tertiary region of the ferret auditory cortex.Nat Neurosci 22:447–459.

Faure PA, Fremouw T, Casseday JH, Covey E (2003) Temporalmasking reveals properties of sound-evoked inhibition in duration-tuned neurons of the inferior colliculus. J Neurosci 23:3052–3065.

Felix RA 2nd, Fridberger A, Leijon S, Berrebi AS, Magnusson AK(2011) Sound rhythms are encoded by postinhibitory reboundspiking in the superior paraolivary nucleus. J Neurosci 31:12566–12578.

Felix RA 2nd, Kadner A, Berrebi AS (2012) Effects of ketamine onresponse properties of neurons in the superior paraolivary nucleusof the mouse. Neuroscience 201:307–319.

Felix RA 2nd, Vonderschen K, Berrebi AS, Magnusson AK (2013)Development of on-off spiking in superior paraolivary nucleusneurons of the mouse. J Neurophysiol 109:2691–2704.

Felix RA 2nd, Magnusson AK (2016) Development of excitatorysynaptic transmission to the superior paraolivary and lateral supe-rior olivary nuclei optimizes differential decoding strategies. Neu-roscience 334:1–12.

Felix RA 2nd, Gourévitch B, Gómez-Álvarez M, Leijon SCM, SaldañaE, Magnusson AK (2017) Octopus cells in the posteroventral co-chlear nucleus provide the main excitatory input to the superiorparaolivary Nucleus. Front Neural Circuits 11:37.

Felix RA 2nd, Gourévitch B, Portfors CV (2018) Subcortical path-ways: towards a better understanding of auditory disorders. HearRes 362:48–60.

Forrest TG, Green DM (1987) Detection of partially filled gaps in noiseand the temporal modulation transfer function. J Acoust Soc Am82:1933–1943.

Gómez-Álvarez M, Gourévitch B, Felix RA 2nd, Nyberg T,Hernández-Montiel HL, Magnusson AK (2018) Temporal informa-



tion in tones, broadband noise, and natural vocalizations is con-veyed by differential spiking responses in the superior paraolivarynucleus. Eur J Neurosci 48:2030–2049.

Grothe B (1994) Interaction of excitation and inhibition in processingof pure tone and amplitude-modulated stimuli in the medial supe-rior olive of the mustached bat. J Neurophysiol 71:706–721.

Guinan JJ Jr, Guinan SS, Norris BE (1972) Single auditory units in thesuperior olivary complex I: responses to sounds and classifica-tions based on physiological properties. Int J Neurosci 4:101–120.

Guzman SJ, Schlögl A, Schmidt-Hieber C (2014) Stimfit: quantifyingelectrophysiological data with Python. Front Neuroinform 8:16.

Helfert RH, Bonneau JM, Wenthold RJ, Altschuler RA (1989) GABAand glycine immunoreactivity in the guinea pig superior olivarycomplex. Brain Res 501:269–286.

Henry FM (1948) Discrimination of the duration of a sound. J ExpPsychol 38:734–743.

Hines ML, Carnevale NT (2001) NEURON: a tool for neuroscientists.Neuroscientist 7:123–135.

Horwitz AR, Ahlstrom JB, Dubno JR (2011) Level-dependentchanges in detection of temporal gaps in noise markers by adultswith normal and impaired hearing. J Acoust Soc Am 130:2928–2938.

Ison JR (1982) Temporal acuity in auditory function in the rat: reflexinhibition by brief gaps in noise. J Comp Physiol Psychol 96:945–954.

Jalabi W, Kopp-Scheinpflug C, Allen PD, Schiavon E, DiGiacomoRR, Forsythe ID, Maricich SM (2013) Sound localization ability andglycinergic innervation of the superior olivary complex persist aftergenetic deletion of the medial nucleus of the trapezoid body. JNeurosci 33:15044–15049.

Kadner A, Kulesza RJ Jr, Berrebi AS (2006) Neurons in the medialnucleus of the trapezoid body and superior paraolivary nucleus ofthe rat may play a role in sound duration coding. J Neurophysiol95:1499–1508.

Kitzes LM, Gibson MM, Rose JE, Hind JE (1978) Initial dischargelatency and threshold considerations for some neurons in cochlearnuclear complex of the cat. J Neurophysiol 41:1165–1182.

Klink KB, Klump GM (2004) Duration discrimination in the mouse(Mus musculus). J Comp Physiol A Neuroethol Sens Neural BehavPhysiol 190:1039–1046.

Klug A, Khan A, Burger RM, Bauer EE, Hurley LM, Yang L, Grothe B,Halvorsen MB, Park TJ (2000) Latency as a function of intensity inauditory neurons: influences of central processing. Hear Res 148:107–123.

Kopp-Scheinpflug C, Tolnai S, Malmierca MS, Rübsamen R (2008)The medial nucleus of the trapezoid body: comparative physiol-ogy. Neuroscience 154:160–170.

Kopp-Scheinpflug C, Tozer AJ, Robinson SW, Tempel BL, HennigMH, Forsythe ID (2011) The sound of silence: ionic mechanismsencoding sound termination. Neuron 71:911–925.

Kopp-Scheinpflug C, Sinclair JL, Linden JF (2018) When soundstops: offset responses in the auditory system. Trends Neurosci41:712–728.

Kulesza RJ Jr (2008) Cytoarchitecture of the human superior olivarycomplex: nuclei of the trapezoid body and posterior tier. Hear Res241:52–63.

Kulesza RJ Jr, Spirou GA, Berrebi AS (2003) Physiological responseproperties of neurons in the superior paraolivary nucleus of the rat.J Neurophysiol 89:2299–2312.

Kulesza RJ Jr, Kadner A, Berrebi AS (2007) Distinct roles for glycineand GABA in shaping the response properties of neurons in thesuperior paraolivary nucleus of the rat. J Neurophysiol 97:1610–1620.

Kuwada S, Batra R (1999) Coding of sound envelopes by inhibitoryrebound in neurons of the superior olivary complex in the unanes-thetized rabbit. J Neurosci 19:2273–2287.

Löhrke S, Srinivasan G, Oberhofer M, Doncheva E, Friauf E (2005)Shift from depolarizing to hyperpolarizing glycine action occurs atdifferent perinatal ages in superior olivary complex nuclei. Eur JNeurosci 22:2708–2722.

Moore BCJ (1997) An introduction to the psychology of hearing, Ed4. San Diego, CA: Academic Press.

Morgan MJ, Thompson P (1975) Apparent motion and the Pulfricheffect. Perception 4:3–18.

Mountcastle VB, Davies PW, Berman AL (1957) Response propertiesof neurons of cat’s somatic sensory cortex to peripheral stimuli. JNeurophysiol 20:374–407.

Oertel D, Cao XJ, Ison JR, Allen PD (2017) Cellular computationsunderlying detection of gaps in sounds and lateralizing soundsources. Trends Neurosci 40:613–624.

Ollo C, Schwartz IR (1979) The superior olivary complex in C57BL/6mice. Am J Anat 155:349–373.

Penner MJ (1977) Detection of temporal gaps in noise as a mea-sure of the decay of auditory sensation. J Acoust Soc Am61:552–557.

Pérez-González D, Malmierca MS, Moore JM, Hernández O, CoveyE (2006) Duration selective neurons in the inferior colliculus of therat: topographic distribution and relation of duration sensitivity toother response properties. J Neurophysiol 95:823–836.

Phillips DP, Hall SE, Boehnke SE (2002) Central auditory onsetresponses, and temporal asymmetries in auditory perception.Hear Res 167:192–205.

Pilati N, Linley DM, Selvaskandan H, Uchitel O, Hennig MH, Kopp-Scheinpflug C, Forsythe ID (2016) Acoustic trauma slows AMPAreceptor-mediated EPSCs in the auditory brainstem, reducingGluA4 subunit expression as a mechanism to rescue binauralfunction. J Physiol 594:3683–3703.

Pollak GD, Gittelman JX, Li N, Xie R (2011) Inhibitory projections fromthe ventral nucleus of the lateral lemniscus and superior paraoli-vary nucleus create directional selectivity of frequency modula-tions in the inferior colliculus: a comparison of bats with othermammals. Hear Res 273:134–144.

Pratt H, Bleich N, Mittelman N (2005) The composite N1 componentto gaps in noise. Clin Neurophysiol 116:2648–2663.

Rhode WS, Oertel D, Smith PH (1983) Physiological response prop-erties of cells labeled intracellularly with horseradish peroxidase incat ventral cochlear nucleus. J Comp Neurol 213:448–463.

Roberts MT, Seeman SC, Golding NL (2014) The relative contribu-tions of MNTB and LNTB neurons to inhibition in the medialsuperior olive assessed through single and paired recordings.Front Neural Circuits 8:49.

Saldaña E, Aparicio MA, Fuentes-Santamaría V, Berrebi AS (2009)Connections of the superior paraolivary nucleus of the rat: projec-tions to the inferior colliculus. Neuroscience 163:372–387.

Sayegh R, Aubie B, Faure PA (2011) Duration tuning in the auditorymidbrain of echolocating and non-echolocating vertebrates. JComp Physiol A Neuroethol Sens Neural Behav Physiol Neuro-ethol Sens Neural Behav Physiol 197:571–583.

Schofield BR (1991) Superior paraolivary nucleus in the pigmentedguinea pig: separate classes of neurons project to the inferiorcolliculus and the cochlear nucleus. J Comp Neurol 312:68–76.

Schofield BR (1995) Projections from the cochlear nucleus to thesuperior paraolivary nucleus in guinea pigs. J Comp Neurol 360:135–149.

Scholl B, Gao X, Wehr M (2010) Nonoverlapping sets of synapsesdrive on responses and off responses in auditory cortex. Neuron65:412–421.

Smith AJ, Owens S, Forsythe ID (2000) Characterisation of inhibitoryand excitatory postsynaptic currents of the rat medial superiorolive. J Physiol 529 [Pt 3]:681–698.

Sohoglu E, Chait M (2016) Neural dynamics of change detection incrowded acoustic scenes. Neuroimage 126:164–172.

Sollini J, Chapuis GA, Clopath C, Chadderton P (2018) ON-OFFreceptive fields in auditory cortex diverge during developmentand contribute to directional sweep selectivity. Nat Commun9:2084.



Steinert JR, Kopp-Scheinpflug C, Baker C, Challiss RA, Mistry R,Haustein MD, Griffin SJ, Tong H, Graham BP, Forsythe ID(2008) Nitric oxide is a volume transmitter regulating postsyn-aptic excitability at a glutamatergic synapse. Neuron 60:642–656.

Steinert JR, Postlethwaite M, Jordan MD, Chernova T, Robinson SW,Forsythe ID (2010) NMDAR-mediated EPSCs are maintained andaccelerate in time course during maturation of mouse and ratauditory brainstem in vitro. J Physiol 588:447–463.

Sun YJ, Wu GK, Liu BH, Li P, Zhou M, Xiao Z, Tao HW, Zhang LI(2010) Fine-tuning of pre-balanced excitation and inhibition duringauditory cortical development. Nature 465:927–931.

Thompson AM, Thompson GC (1991) Posteroventral cochlear nucleusprojections to olivocochlear neurons. J Comp Neurol 303:267–285.

Walcher J, Hassfurth B, Grothe B, Koch U (2011) Comparativeposthearing development of inhibitory inputs to the lateral superiorolive in gerbils and mice. J Neurophysiol 106:1443–1453.

Wehr M, Zador AM (2003) Balanced inhibition underlies tuning andsharpens spike timing in auditory cortex. Nature 426:442–446.

Yassin L, Radtke-Schuller S, Asraf H, Grothe B, Hershfinkel M,Forsythe ID, Kopp-Scheinpflug C (2014) Nitric oxide signalingmodulates synaptic inhibition in the superior paraolivary nucleus(SPN) via cGMP-dependent suppression of KCC2. Front NeuralCircuits 8:65.



Edinburgh Research Explorer · 2019-06-21 · May/June 2019, 6(3) e0106-19.2019 eNeuro.org. NMDA...

Documents

Transcript of Edinburgh Research Explorer · 2019-06-21 · May/June 2019, 6(3) e0106-19.2019 eNeuro.org. NMDA...